Projects:SobolevTracker

Tracking With Adaptive Sobolev Active Contours

In this project we propose adaptive tracking mechanism which can be used in military and civilian surveillance applications as well as in medical video applications, or 3D volume segmentation. The proposed Sobolev active contour model overcomes the problems of occlusions and changes in scale by adaptive tweaking of the rigidity parameters. The proposed tracking algorithms work in a variety of scenarios and deal naturally with clutter and noise in the scenes, object deformations, partial and entire object occlusions, and low contrast objects. Experimental results show the advantages of our approach compared to state-of-the-art visual trackers [1].

In addition, we have proposed an algorithm for active contour self-crossing detection and elimination [2]. The problem of self-crossing is well known for parametric active contours, and it present a very important obstacle for the successful tracking. The proposed solution is based on topological properties of closed contours.

Results

Figures 1-3 show an example of three frames of successful hand tracking, despite of different out of plane deformations, changing motion dynamics, and contour self crossings.

• 

  Figure 1. Tracked frames of video sequence

• 

  Figure 2.

• 

  Figure 3.

Publications

1. A. Nakhmani, A. Tannenbaum, "Tracking with Adaptive Sobolev Snakes." Submitted to IEEE Transactions on Image Processing.

2. A. Nakhmani, A. Tannenbaum, "Self-Crossing Detection and Location for Parametric Active Contours," IEEE Transactions on Image Processing, DOI:10.1109/TIP.2012.2188808, Volume 21, Issue 7, pp. 3150-3156, July 2012.

Key Investigators

• UAB: Arie Nakhmani and Allen Tannenbaum